# Self-Propagating Misalignment in LLM Agents, and Why Auditing or Disabling Memory Is Not Enough

> Полный текст доступен на английском

**Published:** 6 октября 2026 г.
**Author:** Debeshee Das, Jacqueline Tay, Bruce Tsai, David Huang, Javier Rando
**Source:** arXiv — cs.AI
**Language:** en-US
**HTML:** https://aiqazaqstan.com/ru/articles/self-propagating-misalignment-in-llm-agents-and-why-auditing-or-disablin

Memory poisoning attacks on LLM agents typically assume an external adversary who plants content in the agent's persistent memory to steer its behavior. We instead study, with no adversary involved, whether a misaligned agent can write a goal it cannot yet act on to persistent…

> Материал агрегирован из первоисточника: **arXiv — cs.AI**
> [Читать полностью в arXiv — cs.AI](https://arxiv.org/abs/2610.04083)

**Теги:** #arXiv

---

*AI Qazaqstan — non-commercial informational portal about AI in Kazakhstan. Machine translations and summaries are AI-generated.*